IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO.

2, MAY 2013

639

DG Allocation for Benefit Maximization

in Distribution Networks
Mostafa F. Shaaban, Student Member, IEEE, Yasser M. Atwa, Student Member, IEEE, and
Ehab F. El-Saadany, Senior Member, IEEE

AbstractThis paper proposes a method to evaluate the worth

of installing renewable distributed generation (DG) in distribution
networks. Moreover, the work optimally allocates these DG units in
the distribution network to maximize the worth of the connection
to the local distribution company (LDC), as well as the customers
connected to the system. The proposed methodology helps the LDC
to better assess the benefits of the renewable DG units proposed
connections and to identify the optimal buses on which to connect
these DG units. The benefits considered in this paper are deferral
of upgrade investments, reduction of the cost of energy losses, and
reliability improvement, which is represented by the interruption
cost reduction. The proposed methodology takes into consideration
the uncertainty and variability associated with the output power of
renewable DG as well as the load variability. The planning problem
of determining the optimal location and sizes of DG units is defined
as multi-objective mixed integer programming.
Index TermsDistributed generation, distribution system,
Monte Carlo methods, probability density function.

I. INTRODUCTION

ITH the challenges currently facing electricity generation and the recent restructure of energy systems, renewable distributed generation (DG) becomes a vital option.
Renewable energy resources are the only option to a sustainable
energy supply infrastructure since they are neither exhaustible
nor polluting [1]. However, installing DG units in the distribution network has a significant effect, as the network was originally planned for passive circuits. Improper allocation of DG
units in the distribution system might lead to detrimental effects; therefore, the accommodation of a high penetration of DG
units in the power system has to be planned carefully through
allocation of these DG units to maximize their benefits without
violating system constraints. Further, the intermittent nature of
renewable DG units adds more technical and economic challenges to the planning problem.
There are numerous benefits that can be achieved for the local
distribution company (LDC) from the installation of DG units,
among which three are considered in this work. The first benefit
is to relieve the congestions in network feeders and defer the
previously required system upgrades. In [2], a multi-period optimal power flow was used for optimal allocation of DG units in
a distribution network. A methodology was developed in [3] for
Manuscript received July 04, 2011; revised October 11, 2011, January
02, 2012, and May 22, 2012; accepted August 07, 2012. Date of publication September 27, 2012; date of current version April 18, 2013. Paper no.
TPWRS-00590-2011.
The authors are with the Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, ON N2L 3G1, Canada (e-mail:
m2farouk@uwaterloo.ca; yatwa@groupesm.com; ehab@uwaterloo.ca).
Digital Object Identifier 10.1109/TPWRS.2012.2213309

optimal allocation of shunt capacitors to maximize the savings

from reduced losses and reinforcements. Different methodologies were presented for evaluating the effect of DG on reinforcement investment deferral in [4] and [5]. In [6], the authors presented a heuristic approach for the optimal investments of distribution companies to meet their load growth. However, [2][6]
only considered dispatchable DG units in their studies, while in
[7] the system peak load and the capacity credit of wind based
DG units were utilized in the planning problem.
The second benefit of installing DG units in the distribution network is the reduction of energy losses. Some work proposed the optimal placement of DG units for minimizing power
losses in the system as in [8], where a time-varying load and
DG power were considered. In [9], the optimal DG penetration
of different technologies was calculated by examining the effect of changing the penetration level on annual energy losses.
An approach was proposed in [10] for finding the optimal size
and power factor of four types of dispatchable DG units. The
work presented in [11] proposed a multi-objective DG allocation approach to minimize real and reactive power losses. However, the work presented in [8][11]considered only conventional types of DG units with dispatchable sizes. On the other
hand, the work in [1] proposed an approach for optimal renewable DG placement to minimize energy losses, considering the
variability and uncertainty associated with renewable DG resources. In [12], the effect of renewable DG units allocation in
minimizing energy losses was demonstrated through a multi-period AC optimal power flow based technique, taking into consideration smart control schemes. A multi-objective approach
was presented in [13] for optimal allocation of variable and controllable DG units in the system to minimize different objectives
including line losses and emissions. However, reducing the total
annual energy losses is not an accurate representation of LDC
requirements, but the cost of annual energy losses is the factor
which must be considered.
The third benefit is improved reliability of the power supply
for different customers. Most of the previously mentioned publications have not investigated the impacts of DG units on system
reliability. However, there is some work presented in this area,
such as the work in [14], where an optimal placement of DG
units for maximum improvement of system reliability is proposed. In [15], a multi-objective technique was developed for
optimal allocation of DG considering system reliability. An algorithm to determine the optimum operating strategy for DG
was presented in [16] incorporating reliability worth evaluation of distribution system, to minimize the cost of customer interruption. However, the work presented in [14][16] assumed
conventional dispatchable DG.

0885-8950/$31.00 2012 IEEE

640

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

From the above discussion, it is obvious that sufficient work

has been done with respect to the allocation of DG units in
the distribution system; unfortunately, little work has addressed
renewable DG. Moreover, optimizing one objective in the allocation problem could negatively impact other system operational costs. Therefore, the allocation problem must be considered multi-objective in order to assess the benefits of adding renewable DG to the distribution system.
This paper presents a methodology to evaluate the economic
benefits of renewable DG. A genetic algorithm (GA) based
methodology is implemented for optimal DG allocation, to
maximize the deferral of system upgrade investments, reduction of the energy losses cost and reduction of the interruption
cost. The proposed methodology takes into account:
uncertainty and variability associated with DG output;
variable hourly cost of energy;
load variability and customer sector type;
protection and metering equipment upgrades;
nonlinear time dependent cost damage function.
The main contributions of this paper can be summarized as
follows:
The paper presents a methodology that considers the variability and uncertainty of renewable resources in evaluating upgrade costs, where a new factor is defined to represent the lines overloading risk.
Unlike, previous work that focused on minimizing the cost
of the energy losses using fixed energy price, a methodology to evaluate the cost of energy losses in distribution
systems utilizing variable hourly cost of energy is proposed
in this paper.
A methodology to evaluate the cost of interruption in distribution network for long-term planning is presented in
this paper. The methodology takes into consideration stochastic nature of renewable DG and time dependent interruption cost.
The paper is organized as follows: the next three sections
present problem description, modeling, and problem formulation. Section V presents a sample case study of a distribution
system. Results and discussion are presented in Section VI. Finally, the conclusions are presented in Section VII.
II. PROBLEM DESCRIPTION
In this section, the system costs considered in the proposed
long-term planning problem are described.
A. System Upgrade Costs
System upgrade cost in this work is considered as the sum
of lines, protection and metering equipment upgrade costs. It
is assumed that the main substation transformers are redundant,
which is a common practice in Ontario, Canada. The considered
costs are described as follows:
1) Lines reinforcement costs
Due to load growth, lines or cable upgrades may become essential. In addition, lines upgrade can be used
to avoid voltage violation and to increase system security. If carefully planned, installing DG units in a distribution system can relieve congestions on feeders; thus,
it can defer these upgrades.

2) Protection and metering equipment upgrades

Due to high penetration of DG, reverse power flow at
the substation can occur. Accordingly, metering equipment at the substation needs to be upgraded. Moreover,
installing DG units in the system contributes to the short
circuit levels and may require upgrading the protective
equipment.
B. Cost of Energy Losses
Installing DG units in a distribution network affects the
energy losses; however, due to the variability of load, energy
prices and DG units output power, the cost of annual energy
losses has to be calculated hourly. This means that the load
flow analysis has to be performed Y*8760 times, where Y is
the number of scenarios generated by Monte Carlo simulation
(MCS), which is not practical. In this paper, a methodology
is proposed to limit the number of load flow analysis to the
number of states of the combined load and DG model.
Predicting the electric energy prices variations along the
planning horizon is a very complicated process as it depends on
many factors. Thus, the energy prices variations are assumed
to be input to this study and are out of scope. Accordingly, the
hourly prices of energy in Ontario, Canada, 2010 are utilized
to represent each year in the planning horizon. Using variable
hourly prices of energy is assumed to better assess the effect
of renewable DG on system losses. This is because this effect
can be concentrated during low or high energy prices during
the day.
C. Cost of Interruption
The distribution system is an important link between the
transmission-generation system and customers. In most cases,
these links are radial, which makes them susceptible to outage
due to the failures of a single element. Statistics analyzed by
the Canadian Electrical Association indicate that almost 80%
of the outages seen by Canadian utility customers are due to
the distribution system [17]. Typically, the system has two
operating modes:
Grid connected
The grid and the installed DG are supplying the load requirements. Both the dispatchable and the renewable DG
dump their generated power to the system.
Islanded system
A distribution network is fed from a transmission network,
and when the connection to the transmission system is lost,
the distribution network is islanded. According to [18],
DG units can supply the system loads during scheduled
or unscheduled outage events, which can improve system
reliability.
The following assumptions are considered in the proposed
methodology to evaluate the cost of interruption in the distribution networks with renewable DG connections.
Islanding mode of operation is assumed to be allowed [18],
which benefits both the LDC and the customers.
During islanded model of operation, relying on renewable
DG only may cause stability problems regarding voltage
and frequency variations [19]. Accordingly, dispatchable

SHAABAN et al.: DG ALLOCATION FOR BENEFIT MAXIMIZATION IN DISTRIBUTION NETWORKS

DG units are typically responsible of managing the power

balance in the island between generation (involving dispatchable and renewable DG units) from one side and the
load from other side. Hence, a minimum percentage of dispatchable DG has to be assumed to assure successful islanding if there is available power. As given in [19], the
threshold is assumed to be 60% of the total DG units installed in the system. Below this threshold, the wind-based
DG units are required to disconnect during islanding mode
of operation.
For an island to be successful, the generated power from all
DG units within the island should be higher than or equal
to a certain percentage of load required power. There is
no upper limit to the connected DG units within the island
to be successful, as it is assumed that communication signals control the output of dispatchable DG units, and if required, curtail the output of renewable DG units.
III. GENERATION AND LOAD MODELING
In this section the generation and load modeling are described, where the following assumptions are made.
Hourly average load and wind speed data are considered in
this work and the variations within the hour are neglected.
Wind DG output power and load are modeled as a multistate variables, where the number of states represents a
trade-off between accuracy and complexity of the planning
problem [20].
A. Wind Generation Modeling
MCS is utilized for evaluating the cost of energy losses and
cost of interruption, due to the variable hourly cost of energy
and the non-linear cost damage function. On the other hand, a
probabilistic wind speed model is used for evaluating the costs
of upgrade. Thus, two models are described for the wind DG
output power, probabilistic model and MCS model as follows:
1) Divide the entire year into clusters (seasons or months).
2) Generate a typical 24 hours day for each cluster in order to
represent the random behavior of the wind speed.
3) Model the renewable resource behaviour during each
hour by a proper probability density function (pdf), then
generate the probabilistic wind speed model utilizing the
actual recorded historical wind speed data. A Weibull
pdf is the most commonly used pdf to describe the wind
variability [21]. Its formula, given in [22], depends on two
parameters to fit the distribution function to the measured
values of wind speeds. The parameters are calculated
using the mean wind speed and the standard deviation as
in [22].
4) Use the inverse of the cumulative distribution function of
each hour, as given in [23], in MCS to generate virtual scenarios of the hourly wind speed, where a proper stopping
criterion is required according to the type of analysis.
5) Utilize the failure rate of the wind based DG to generate an
artificial two state availability model.
6) Convolve the models obtained in steps 4 and 5 to generate
the final MCS model.

641

7) Divide the entire speed range into a proper number of states

according to the accuracy and speed of the simulation.
8) Calculate the output power of the wind based DG unit corresponding to the wind speed states in the wind DG models
using the formula in [1].
B. Dispatchable DG Unit Modeling
Dispatchable DG units can be divided into two groups: synchronous machine based (as diesel and natural gas based DG)
and inverter based (as fuel cell and micro turbine based DG). In
this work, natural gas DG units are considered. The output of
these DG units is assumed to be fixed in normal mode of operation. However, during islanding mode the output of these DG
units is assumed to be varied to manage the active and reactive
power balance. A two-state-model is used to model the operation of each DG [24]. This model is used in MCS to produce
an artificial annual operating scenario for each dispatchable DG
unit.
C. Load Modeling
The load in the distribution network under study is assumed
to follow the IEEE reliability test system load pattern [25]. The
load is modeled by a definite number of states depending on desired accuracy, time scale and speed of simulation. Uncertainty
of certain percentages could be used to generate different annual
scenarios.
IV. PROBLEM FORMULATION
In this section the proposed DG planning problem formulation is presented, which is classified as mixed integer nonlinear
programming. The following assumptions are made.
Most of the utilities force the DG units to operate in constant power factor mode. Thus, the DG units are assumed
to operate at unity power factor [1].
DG units capacities are discretized at a definite step, which
is assumed to be 100 kW in the presented work.
For combining the effect of DG units installation on system
upgrade, energy losses and reliability, the typical costs in Canadian dollars are used for each individual objective. In the next
section GA is utilized to find the optimal sizes and locations
of DG units to minimize the objective function. The proposed
planning problem is described by the following.
Objective function:
Minimize:

(1)
where is a binary variable corresponding to constraint (the
second term represents a penalty factor for violating constraint
); nc is the total number of constraints.
The incentives here are assumed to be a money value received by the LDC for each renewable MW connected to the
system. They are not considered in the presented results in
the next section. The cost(s) of objective(s) in (1) can be the

642

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

individual cost or sum of different costs as described in next

subsections.
Subject to:
Power flow constraints:

(2)
Fig. 1. Structure of a typical chromosome in the proposed planning problem.

(3)
where and are the bus number; is the total number of
buses in the system under study; is the state number; is
and
are the active and reactive
the year under study;
power demands;
and
are the active and reactive generated powers.
Voltage limits constraints:
(4)
Maximum penetration:
Maximum penetration is taken so as to limit maximum reverse power flow at 60% of substation rating during minimum
load condition [26]:
(5)
,
, and
are the generated power from
where
dispatchable DG units, wind DG units and main substation, respectively.
Discrete size of DG units:
(6)
(7)
are binary variwhere and are integer variables; and
ables indicating the decision of installing dispatchable DG unit
and wind based DG unit at bus , respectively.
Candidate buses:
(8)
and
where
respectively.
DG units limit

are sets of all buses and candidate buses,

(9)
and
are the maximum number of DG units inwhere
stalled in the system for dispatchable and wind based DG, respectively.
The meta-heuristic optimization techniques family proved its
effectiveness in solving many complicated practical problems,

such as DG planning [13], unit commitment [27], and economical dispatch [28]. In the planning problem presented in this
paper the GA is utilized, which is a population based searching
algorithm. The population consists of chromosomes, and each
chromosome consists of a number of genes.
For radial distribution systems the number of genes is selected to be a multiple of the number of candidate buses according to the types of DG units to be installed. For example,
in case of considering two types of DG units in the allocation
problem, like natural gas and wind based DG, each chromosome
in the population consists of a vector of length equals to four
times the number of the candidate busses, as shown in Fig. 1.
For each candidate bus, there are four genes, which are shaded
in Fig. 1. Two genes carry binary values, which indicate the decision of installing DG units for the two DG types at the corresponding bus. The other two genes carry integer values, which
indicate the capacity of the corresponding DG units as a multiple of a definite step for each DG type.
On the other hand, for mesh distribution networks, where
there is more than one decision for lines upgrades, the chromosome should include two extra genes for each line in the system,
as shown in Fig. 1. The first gene carries a binary value that represents the decision of upgrading the corresponding line. The
second gene carries an integer value that represents the year of
upgrade.
In each iteration, the fitness of each individual in the population is evaluated. This fitness is the total cost of the considered
objective(s) in the objective function described in (1). These
costs are described in the following subsections.
A. System Upgrades
This subsection describes the methodology proposed for
evaluating the cost of system upgrades. A risk factor (RF)
is proposed which represents the expected duration of overloading per year. This factor is used in evaluating the cost of
lines upgrades.
1) Lines upgrades
For radial systems, considering the no DG case, the
reinforcement costs can be evaluated at the extreme
condition of power flow in the lines, which is simply
one condition at peak load, as the power flow is always
from the substation to the load points. However, in the
case of DG units presence in the system, the load flow
analyses is performed for each state for this study. The
wind based DG is modeled by 14 states model [1] and

SHAABAN et al.: DG ALLOCATION FOR BENEFIT MAXIMIZATION IN DISTRIBUTION NETWORKS

the load is modeled by 8 states model. The procedure

for evaluating the cost of system upgrades is described
as follows:
a) For each state , go through steps b to d.
b) For each year , go through steps c to d.
c) Update loads with annual rise, and run load flow
analysis for state and year .
d) For each line, record the year in which upgrade
is required and calculate the corresponding net
present value (NPV) of the cost of upgrade of each
line for each state .
e) For each line, arrange the combined generation and
load states in descending order according to the
calculated NPV.
f) For each line, if the probability of the state corresponding to the maximum cost of upgrade is above
the RF, proceed to step g; if not, proceed to the next
state. If the sum of the probabilities of this state and
the previous state(s) is higher than the RF go to step
g; if not, proceed to next state and repeat the previous statement.
g) Record this upgrade cost and repeat step f for the
next line.
h) The NPV of the required reinforcement investments during the period under study can be evaluated using the following formula [6]:
(10)
is the NPV of the reinforcement
is
where
is the discount rate;
the cost of the reinforcement
is the total number of required reinforcements; is the
year when the reinforcement is required.
For evaluating the cost of lines upgrade in case of wind
based DG units, if the RF is taken to be zero, the cost is
greater than or equal to the cost of lines upgrade without
DG. For example, assume the combined load and generation states are 224 states. These states have two extremes, defined as rated DG output power with minimum
load and zero DG output power with peak load. Each
state contributes to line X upgrade. Assume that the state
which contributes the most has a probability of occurrence of 0.05%. This corresponds to almost 4 hours per
year. For zero RF, all the states are considered including
the second extreme mentioned above, which is the case
without DG. Thus, for zero RF, the cost of each line upgrade is greater than or equal to the case without DG.
For an RF of 6 hours per year, the most contributing state
will be neglected because its probability of occurrence
is less than the RF, and other states will be considered.
This may result in reduced cost of the lines upgrade than
the base case.
2) Metering equipment upgrade
At the substation terminals where the metering devices
are installed, the direction of power flow is checked
under the state of minimum load and rated DG output.

643

Accordingly, the cost of upgrading the metering devices

is evaluated.
3) Protection switch gear upgrade
To prevent false tripping and for effective fault clearing,
a short circuit analysis of the system has to be carried out
in the presence of installed DG units. Accordingly, the
cost of upgrading the protective equipment is evaluated.
B. Cost of Energy Loss
The power loss for each state of the combined generation and
load states is evaluated for 20 years with load growth. Then,
the cost of the annual energy losses is evaluated for each year
according to the following methodology.
The power loss for each year is represented as a vector of
in which each element represents the power loss
length
corresponding to state :
(11)
A binary variable is defined as
(12)
where
is the total number of states of the combined load and
generation model;
is the total number of scenarios in the
probabilistic chronological model.
For the binary variable in (12), each row consists of
zeros and one element of value 1; this element corresponds
to the actual load state. This variable is generated only once
for certain wind or solar regimes, and it allows for the hourly
evaluation of the cost of energy losses.
For example, assume that the load states are given by
and the generation states are given by
. Thus, there are
4 combined load and generation states, which are given by:
. Assume that the period under study is 5 hours. The hourly load curve and the hourly
DG output are given by:
,
; this means that in the first time segment the second
state (0.5,1) occurs, then the first state (0.5,0) occurs and so on.
The state number represents the locations of the ones in the rows
of the binary variable which is given by

(13)

The cost of annual energy losses is evaluated by

(14)
is the cost of annual energy losses for year y.
where
Vector represents the hourly energy price in $/kWh for the
8760 hours. The hourly market clearing prices of electric energy
in 2010 from the IESO website [29] are utilized as vector .

644

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

Fig. 2. CDF of different customers [30].

Finally, the NPV of the total cost of energy losses for the period
under study is evaluated by
(15)

C. Cost of Interruption
The distribution network usually contains a mix of residential, commercial and industrial customers. According to [29],
the cost of interruption, which is known as the cost of damage
function (CDF), is not linear and varies according to the duration of interruption as shown in Fig. 2[30], which shows the
average cost of interruption estimates obtained as a function of
interruption duration for each customer sector.
Since the CDF is not linear, as seen in Fig. 2, the outage
cost cannot be evaluated analytically; thus, MCS is utilized. The
outage cost is evaluated using
(16)
where
is the outage cost of the load point
is the outage cost corresponding to interruption event
is
total number of interruption events for load
is the load
point average demand power.
In the above mentioned method to evaluate the contribution
of DG to the interruption cost of different customers, the CDF
is assumed to be constant for certain duration of outage. For
example, the cost for interruption of 2 hours is the same for a
certain customer type whenever this 2 hour interruption occurs
during the day. To accurately express the effect of these DG
units to reduce the cost of interruption, a modification of the
CDF to be dependent on the time of the interruption event is
assumed. For example, an interruption event of 4 hours at peak
load costs more than the similar event at minimum load. Accordingly, the cost of interruption for a certain load point could
be evaluated using

(17)

where
is the time in hours for outage event
is the
per unit load power at time for load point .
The procedure to evaluate the cost of interruption is described
as follows:
1) Divide the system into segments, according to the locations
of protective devices as in [14].
2) For each segment, perform the following steps.
3) Define two sets: set (1), which includes all elements outside
the segment whose failures cause power outage to the segment, and set (2), all elements within the segment whose
failures cause outage to all load points within the segment.
4) Generate a two state model for each element within the two
sets using the failure rate and repair time of each element,
and combine these models to construct a two state model
for each set.
5) Repeat steps 6 to 10 for each year in the period under study.
6) If there are no dispatchable DG units installed within the
segment, go to step 9.
7) If the percentage of dispatchable DG units within the segment is below 60%, deactivate all renewable DG units
within the segment.
8) For each outage event in set (1), check whether the islanding is successful or not. Yet, there is no standard for the
required reserve margin within the island, as it depends on
load variability, load magnitude, reliability requirements,
types and availability of DG units [18]. Therefore, it is assumed that the island is successful if the sum of the generated output power from all DG units within the segment is
higher than or equals to a certain percentage of the load required power, which represents load requirements, system
losses, and reserve margin. This percentage is assumed to
be 110% for dispatchable DG units only, and 115% in case
of renewable and dispatchable DG units connected to the
system, as excess reserve margin is required due to variability and uncertainty of renewable DG units. Accordingly, the outage event in set (1) is modified to up time for
successful islanding or left as down time for unsuccessful
islanding.
9) Generate the final availability model for the segment under
study by the convolution of set (1) and set (2).
10) Evaluate the cost of interruption of using (17).
11) Evaluate the NPV of the cost of interruption using (18):

(18)

V. SAMPLE CASE STUDY

Consider the distribution system under study in [31] which
contains a mix of residential, commercial and industrial customers being supplied from a common supply point, which is
similar to the Canadian distribution as shown in Fig. 3. The
system data and type of customers are available in [31]. The
total system peak load is 4.37 MVA divided into five segments
[14]. Candidate DG bus locations are determined by detailed
planning analysis including technical, environmental and economic studies, which is assumed as an input and are beyond the

SHAABAN et al.: DG ALLOCATION FOR BENEFIT MAXIMIZATION IN DISTRIBUTION NETWORKS

645

TABLE I
DIFFERENT SCENARIOS

Fig. 3. System under study.

Fig. 4. Results of different scenarios.

scope of the work presented in this paper. The candidate buses

selected in the presented case study are totally arbitrary and are
located as shown in Fig. 3. According to the location of the candidate buses, islanding is effective in reducing the cost of interruption only for segments 3 and 5.
For the purpose of technical evaluation of the DG units effect on reliability, the expected energy not served (EENS) of the
system is evaluated as given in [23]. Maximum number of DG
units in the system is limited to 5 units for each type of DG, as
described in (9).
Four different cases are presented in this work representing
base case, dispatchable DG, wind based DG, and mix of the two
DG types respectively. Each case includes different scenarios.
The objective and the RF for each scenario are shown in Table I.
VI. RESULTS AND DISCUSSIONS
The outcomes of the allocation problem for a 20-year study
period are shown in Fig. 4 and the details are shown in Table II
based on typical prices in Canadian dollars, for each scenario in
Table I.
A. Base Case Results
For the base case defined by scenario (A.0), the total cost of
system upgrade, energy losses and interruption is $2 265 683.
The system upgrade share is 59.2%, while the energy losses

share is 10.4% and the interruptions share 30.4%. These costs

are system dependent, and show that upgrade cost has the
highest share; however, the effectiveness of each cost will
depend on the savings as described in the next cases.
B. Dispatchable DG Results
The next four scenarios present the results obtained for allocating dispatchable DG units in the system under study denoted by case (B), and the objective of each scenario is shown
in Table I. In scenario (B.1), the result was found to be at the
boundary of upgrading the protection and metering equipment,
as increasing any DG unit size beyond the obtained values triggers this equipment upgrade, which increases the total cost of
the required upgrades. The reduction in the cost of upgrades is
77.4%, which is very significant. The saving in the cost of upgrades is due to deferring most of the lines upgrades to further
years. For example, line between buses 3 and 23 in the base case
required to be upgraded at the 3rd year; however, in scenario
(B.1) it is deferred to the 19th year. The cost of energy losses is
reduced by 33.7%, which represents a good result, though it is
not the objective in this scenario. On the other hand, the cost of
interruption is not affected. This is because the DG units 36, 37,
and 38 only affect the cost of interruption in segments 3 and 5
as shown in Fig. 3. Furthermore, the outcomes of scenario (B.1)

646

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

TABLE II
DETAILED RESULTS OF DIFFERENT SCENARIOS

show small capacities at these locations. These capacities do not

affect the interruptions as shown in the results. The total saving
is 49.3%, which proves that the proposed methodology can reduce system costs significantly.
In scenario (B.2), the cost of energy losses is reduced by
49.8%, and the cost of upgrades is reduced by 45.1%, while
the cost of interruption is slightly affected. The results show a
slight reduction in the cost of interruption; this is due to higher
capacities of DG units in segments 3 and 5 than the previous
scenario. However, the reduction in cost of interruption is still
almost negligible, as this is not the objective in this scenario.
The total saving is 32.3%, which shows that the upgrade cost is
more effective as the total saving is better in scenario (B.1).
For scenario (B.3), since the DG units in segments 3 and 5
have technical limitations, and the DG units in other segments
have no effect on the cost of interruption, there may be infinite
solutions that give the minimum cost of interruption. Thus, the
presented solution is the minimum penetration solution. Consequently, the outcomes of the planning problem show installed
DG units only in segments 3 and 5. It is important to observe that
the cost of interruption of segments 3 and 5 cannot be reduced
to zero, as only outages due to failures of set (1) elements are
affected by islanding, while outages due to set (2) elements failures cannot be avoided, as described in Section IV. The savings
in the cost of interruption is 18.2% corresponding to a reduction
in EENS from 13,855 to 11,561 kWh/yr. Meanwhile, the sav-

ings in the costs of upgrades and energy losses are 14.4% and
28.3%, respectively. It is further noted that any increase in the
DG units sizes do not affect the cost of interruption, as no further reduction can be achieved. The total saving is about 17%
which shows that the cost of interruption is the least effective
cost in the case under study. Indeed, these results are system
dependent.
In scenario (B.4), the objective is to minimize the three considered costs. The results are the same as scenario (B.1). This
shows that for the case under study, the upgrades cost is the most
effective cost. However, if the discrete sizes constraints in (6,7)
are relaxed, scenario (B.4) introduces a slightly better results
than scenario (B.1).
C. Wind Based DG Results
Only wind based DG units are considered in this case, so
there is no effect on the cost of interruption. This is because,
the percentage sharing of dispatchable DG units in the system
is lower than the 60% threshold, as explained in Section IV.
For the case under study, the outcomes of the allocation
problem for the cost of upgrade at zero RF are found to converge to the base case without DG units. Zero RF means that
all combined wind and load states are considered regardless of
their probabilities, as the LDC is taking no risk of overloading
their lines. Based on this result, the effect of variability and
uncertainty of wind based DG result in equal or higher upgrades

SHAABAN et al.: DG ALLOCATION FOR BENEFIT MAXIMIZATION IN DISTRIBUTION NETWORKS

647

TABLE III
SCENARIO (C.2) RESULTS USING VARIABLE
AND FIXED HOURLY ENERGY PRICES

Fig. 5. Variation of upgrade costs with RF.

costs in distribution networks compared to the case without

DG, as explained before in Section IV.
Accordingly, in next scenarios, which characterized by nonzero RF, there is a risk of overloading defined by the RF. This
risk arises from neglecting one or more states that contribute
most in the reinforcement requirements. For this study, the RF
was taken with different values (scenario C.1.a and C.1.b) as
seen in Table I. In scenario (C.1.a), the RF value is taken to be 3
hours per year. The outcomes of the allocation problem show a
savings of 8.5% for upgrade costs, while the savings in the cost
of energy losses is 13.1%. In scenario (C.1.b), still the objective
is minimizing the upgrade costs. The RF value is taken to be 6
hours per year; the cost of upgrade was further reduced with a
saving of 27.11%. Further, it was observed that as the RF increases, the cost of upgrading keeps decreasing. In other words,
as the LDC increases the risk of overloading their lines due to
stochastic nature of renewable DG units, the expected upgrade
costs keep decreasing as shown in Fig. 5, where the results of
scenario C.2 are repeated with different values of RF. Moreover,
it is shown that the reduction in the upgrade costs is not uniform,
which is dependent on the wind regime, load curve, and system
under study. On the other hand, if the renewable DG units are
represented with a constant output power DG, the outcomes of
the allocation problem are considered to be misleading results,
as the inherent risk of overloading system lines is not revealed.
Therefore, for the sake of comparison, the capacity credit for
wind DG units in [7] is used to evaluate the upgrade costs. As
shown in Fig. 5, modeling the wind DG units with 24% capacity
credit is equivalent to almost a risk of 30 hours of overloading
per year. These results prove the importance of the proposed
methodology to better assess system upgrades with renewable
DG connections. Thus, the LDC can take the risk defined by the
RF, which allows the renewable DG to contribute in reducing
systems upgrade costs with predefined risk.
In scenario (C.2), the cost of energy losses is reduced by
34.7%, while the cost of upgrade is higher than the base case
by 36.6% with RF of 3 hours per year. The increase in the upgrade costs is due to the objective in this scenario, which is minimizing the energy losses only. By comparing scenario (C.2) to
scenario (B.2), it is clear that the variability and uncertainty of
wind based DG result in limiting the reduction in energy losses
compared to dispatchable DG units.

Furthermore, to show the importance of considering the variable hourly cost of energy in renewable DG planning problem,
the results of scenario (C.2) are compared to the previously proposed approaches in [8][11] and [1], which traditionally minimized the energy losses or the cost of the energy losses using
fixed energy price. As shown in Table III, the obtained results
using the variable hourly cost of energy show NPV of $153 142
for the energy losses. On the other hand, the NPV of the cost of
energy losses for the same DG configuration of scenario (C.2)
was found to be $145 653, using average cost of energy. This
result is lower than the case with variable cost case by 4.9%,
which proves that assuming average cost of energy in renewable DG planning problems may result in inaccurate outcomes,
as renewable DGs effect on system losses is not uniform over
time, and may be concentrated during low or high energy cost
periods.
In scenario (C.3), the total savings in the costs of upgrades
and energy losses is 6.48%, which is slightly higher than scenario (C.1.a), with share of 4.66% for the savings in the cost of
upgrade, and 1.82% for the savings in the cost of energy loss.
Comparing the outcomes of different scenarios in case (C), they
show that, for wind based DG units, the most effective cost is
the cost of energy losses, even if there is a risk of overloading
system lines. The stochastic nature of these renewable DG units
limits their ability of reducing the cost of required upgrades and
energy losses; however, the effect of upgrade costs on the total
savings is more significant due to the RF.
D. Dispatchable and Wind Based DG Results
In case (D), dispatchable and wind based DG units are considered. Due to superiority of dispatchable DG over wind based DG
in minimizing the considered costs, the optimization outcome
is directed to case (B), only deploying dispatchable DG units.
Therefore, to force renewable DG integration, a minimum percentage of 40% of renewable installed DG capacity is assumed.
In other words, the wind DG units capacities should be at least
two-thirds the dispatchable DG units capacities installed in the

648

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

system. Accordingly, the following constraint is added to previously mentioned constraints in Section IV:
(19)
Another way to increase wind based DG penetration is by
adding a monetary value incentive in the objective function in
(1) to mimic the actual incentives the LDC receives from the
government for each MW renewable DG capacity installed in
their system. In scenario (D.1), the savings in the cost of upgrades is 70.7%, which is lower than scenario (B.1) due to green
energy constraint in (19). The total capacity of the wind DG
units is 0.6 MW, which is exactly two-thirds of the total capacity
of the dispatchable DG units. This result shows that the system
is at the boundary of the constraint introduced in (19). In fact, if
this constraint is relaxed, the outcomes of the planning problem
show only dispatchable DG units as mentioned before. In scenario (D.2), the reduction in the cost of energy losses is 49.4%
which is close to the saving in scenario (B.2), but with higher
DG capacities connected to the system. This result shows that a
mix of dispatchable and renewable DG units in the system can
have almost same effect on annual energy loss, but with higher
capacities in case of considering the two types of DG. These
higher capacities are due to the capacity factor of the renewable
DG units [1].
In scenario (D.3) the objective function is the cost of interruption. Although, the total DG capacity connected to the system
in scenario (D.3) is higher than that in scenario (B.3), the maximum saving in the interruption cost is lower. This is due to the
intermittent effect of wind based DG, which limits their contribution in reducing the interruption cost.
Furthermore, to show the difference between using time independent CDF in (16), and time dependent CDF in (17), the same
DG configuration of scenario (D.3) is used to calculate the NPV
of interruption cost using time independent CDF, as described
in Section IV but using (16) instead of (17). The NPV of the
interruption cost is $599 503 for scenario (D.3) using time dependent CDF, as shown in Table II. On the other hand, for same
DG configuration, the cost of interruption was found to be lower
by only 1%, using time independent CDF. This is due to the effect of wind DG on interruption cost, which is concentrated on
low load requirement periods. The difference between the two
methods is not significant in the introduced sample case due to
two reasons: 1) the cost of interruption in the system under study
is not significant; 2) the DG configuration used in comparison
tends to saturate the interruption cost, which is minimized, according to the objective of scenario (D.3). Although, the difference between the two methods is almost negligible, it might be
significant on other systems. Thus, it is recommended to use the
time dependent CDF, which is assumed to result in better assessment of DG impacts on interruption cost.
In scenario (D.4), in which the objective function contains all
the three costs mentioned before, the outcomes are the same as
scenario (D.1), which demonstrates that the savings in the cost
of upgrade is still the most effective.
Comparing scenario (D.4) with total saving of 46% to scenario (B.4) with total saving of 49.3%, the saving when using

a mix of dispatchable and wind based DG units is less than

but comparable to the case of using dispatchable DG units
only. However, in scenario (B.4) there is no risk of overloading
system lines, while in scenario (D.4) there is a risk of 3 hours
of overloading per year.
Finally, the results presented in this section show the effectiveness of the proposed methodology in maximizing the benefits of renewable DG in distribution networks. Moreover, the
methodology introduces more accurate evaluation for the worth
of the renewable DG connection.
VII. CONCLUSION
In this paper, a multi-objective optimization approach based
on GA for optimal allocation of different types of DG units
into the distribution system is proposed. The main objective of
the optimization is to maximize the savings in system upgrades
investment deferral, cost of annual energy losses, and cost of
interruption. The benefits of DG connection are represented
in money value to facilitate comparison and to avoid using
weighting factors, which are usually questionable and may lead
to misleading results.
The proposed method is based on generating probabilistic
and MCS models for the combined generation-load model, including all possible operating conditions. The problem is multiobjective, where typical costs of each objective are represented
in dollars.
The uncertainty of the renewable DG units output is taken
into consideration, as well as, load types and load variability.
The systems technical constraints, protection equipment upgrade, metering equipment upgrade, and different customers
interruption costs are all considered.
Moreover, this work presents a new approach for evaluating
the upgrade requirements in presence of renewable DG in the
distribution systems, where a new factor is introduced to represent the risk of overloading system lines.
In addition, the work presents a new approach to incorporate variable hourly cost of energy along the planning horizon,
while limiting the computational effort to the number of combined generation-load model states. This method is assumed to
represent more accurate evaluation of the cost of energy losses
in long-term planning problems, especially with renewable DG
connections.
Due to the variation of the time of outage events occurrence,
a time dependent CDF is utilized in this work to give more accurate results for the interruption cost, as the effect of renewable DG units during islanding mode may vary according to
the time of outage events occurrence. The proposed planning
technique has been applied to different scenarios for a typical
rural distribution system. The results reveal the effectiveness of
the proposed long-term multi-objective allocation algorithm in
significantly reducing the mentioned costs, taking into consideration all aspects as possible that affect the considered costs,
which benefits both the LDC and the consumers. Using GA allows a satisfactory feasible solution of the planning problem in
a timely manner for the LDC, identifying the best locations in
the system for connecting different types of DG units. The proposed method can be easily applied to any radial system and

SHAABAN et al.: DG ALLOCATION FOR BENEFIT MAXIMIZATION IN DISTRIBUTION NETWORKS

different incentives could be added to the planning problem according to the LDC. Moreover, the proposed algorithm can be
applied to any type of DG units. Dispatchable DG units are dealt
with in the same manner as natural gas DG units, and renewable
DG units are dealt with in the same manner as wind based DG
units. However, some differences may arise that must be taken
into consideration regarding the modeling of each DG type.
REFERENCES
[1] Y. M. Atwa, E. F. El-Saadany, M. M. A. Salama, and R. Seethapathy,
Optimal renewable resources mix for distribution system energy loss
minimization, IEEE Trans. Power Syst., vol. 25, no. 1, pp. 360370,
Feb. 2010.
[2] A. Piccolo and P. Siano, Evaluating the impact of network investment deferral on distributed generation expansion, IEEE Trans. Power
Syst., vol. 24, no. 3, pp. 15591567, Aug. 2009.
[3] H. M. Khodr, F. G. Olsina, P. M. De Oliveira-De Jesus, and J. M. Yusta,
Maximum savings approach for location and sizing of capacitors in
distribution systems, Elect. Power Syst. Res., Elsevier, vol. 78, no. 7,
pp. 11921203, Jul. 2008.
[4] G. Pudaruth and F. Li, Costs and benefits assessment considering deferral of assets expansion in distribution systems, in Proc. 42nd Int.
Universities Power Eng. Conf., UPEC 2007, 2007, pp. 872878.
[5] Y. Zhang, C. Gu, and F. Li, Evaluation of investment deferral resulting
from microgeneration for EHV distribution networks, in Proc. IEEE
Power and Energy Society General Meeting, 2010, pp. 15.
[6] W. El-Khattam, K. Bhattacharya, Y. Hegazy, and M. M. A. Salama,
Optimal investment planning for distributed generation in a competitive electricity market, IEEE Trans. Power Syst., vol. 19, no. 3, pp.
16741684, Aug. 2004.
[7] D. T. C. Wang, L. F. Ochoa, and G. P. Harrison, DG impact on investment deferral: Network planning and security of supply, in Proc.
IEEE Power and Energy Society General Meeting, 2010, p. 1.
[8] C. Wang and M. H. Nehrir, Analytical approaches for optimal placement of distributed generation sources in power systems, IEEE Trans.
Power Syst., vol. 19, no. 4, pp. 20682076, Nov. 2004.
[9] V. H. M. Quezada, J. R. Abbad, and T. G. S. Roman, Assessment
of energy distribution losses for increasing penetration of distributed
generation, IEEE Trans. Power Syst., vol. 21, no. 2, pp. 533540, May
2006.
[10] D. Q. Hung, N. Mithulananthan, and R. Bansal, Analytical expressions for DG allocation in primary distribution networks, IEEE Trans.
Energy Convers., vol. 25, no. 3, pp. 814820, Sep. 2010.
[11] L. F. Ochoa, A. Padilha-Feltrin, and G. P. Harrison, Evaluating distributed generation impacts with a multiobjective index, IEEE Trans.
Power Del., vol. 21, no. 3, pp. 14521458, Jul. 2006.
[12] L. F. Ochoa and G. P. Harrison, Minimizing energy losses: Optimal
accommodation and smart operation of renewable distributed generation, IEEE Trans. Power Syst., vol. 26, no. 1, pp. 198205, Feb. 2011.
[13] A. Alarcon-Rodriguez, E. Haesen, G. Ault, and J. Dries, Multi-objective planning framework for stochastic and controllable distributed
energy resources, IET Renew. Power Gen., vol. 3, no. 2, pp. 227238,
Jun. 2009.
[14] Y. Attwa and E. El-Saadany, Reliability based analysis for optimum
allocation of DG, in Proc. IEEE Canada Elect. Power Conf., EPC
2007, 2007, pp. 2530.
[15] G. Celli, E. Ghiani, S. Mocci, and F. Pilo, A multiobjective evolutionary algorithm for the sizing and siting of distributed generation,
IEEE Trans. Power Syst., vol. 20, no. 2, pp. 750757, May 2005.
[16] I.-S. Bae, J.-O. Kim, J.-C. Kim, and C. Singh, Optimal operating
strategy for distributed generation considering hourly reliability
worth, IEEE Trans. Power Syst., vol. 19, no. 1, pp. 287292, Feb.
2004.
[17] Canadian Electricity Association. [Online]. Available: http://www.
electricity.ca.
[18] IEEE Guide for Design, Operation, and Integration of Distributed Resource Island Systems with Electric Power Systems, IEEE Std. 1547.4,
2011.
[19] R. Karki and R. Billinton, Cost-effective wind energy utilization for
reliable power supply, IEEE Trans. Energy Convers., vol. 19, no. 2,
pp. 435440, Jun. 2004.

649

[20] Y. M. Atwa and E. F. El-Saadany, Reliability evaluation for distribution system with renewable distributed generation during islanded
mode of operation, IEEE Trans. Power Syst., vol. 24, no. 2, pp.
572581, May 2009.
[21] G. Boyle, Renewable Energy: Power for a Sustainable Future. Oxford, U.K.: Oxford Univ. Press, 1996.
[22] S. H. Jangamshetti and V. G. Rau, Site matching of wind turbine generators: A case study, IEEE Trans. Energy Convers., vol. 14, no. 4,
pp. 15371543, Dec. 1999.
[23] R. Billinton and W. Li, Reliability Assessment of Electric Power Systems Using Monte Carlo Methods. New York: Plenum, 1994.
[24] Y. Hegazy, M. Salama, and A. Chikhani, Adequacy assessment of
distributed generation systems using Monte Carlo simulation, IEEE
Trans. Power Syst., vol. 18, no. 1, pp. 4852, Feb. 2003.
[25] IEEE RTS Task Force of APM Subcommittee, The IEEE reliability
test system-1996, IEEE Trans. Power Syst, vol. 14, no. 3, pp.
10101020, Aug. 1999.
[26] HydroOne, Technical DG Interconnection Requirements of Hydro
One. [Online]. Available: http://www.hydroone.com.
[27] S. A. Kazarlis, A. Bakirtzis, and V. Petridis, A genetic algorithm solution to the unit commitment problem, IEEE Trans. Power Syst, vol.
11, no. 1, pp. 8392, Feb. 1996.
[28] C. Chiang, Genetic-based algorithm for power economic load dispatch, IET Gen., Transm., Distrib., vol. 1, no. 2, pp. 261269, Mar.
2007.
[29] [Online]. Available: http://www.ieso.ca.
[30] G. Tollefson, R. Billinton, G. Wacker, E. Chan, and J. Aweya, A Canadian customer survey to assess power system reliability worth, IEEE
Trans. Power Syst., vol. 9, no. 1, pp. 443450, Feb. 1994.
[31] D. Singh and R. Misra, Effect of load models in distributed generation
planning, IEEE Trans. Power Syst., vol. 22, no. 4, pp. 22042212,
Nov. 2007.

Mostafa F. Shaaban (S11) was born in Virginia,

USA, in 1982. He received the B.Sc. and M.Sc.
degrees from Ain Shams University, Cairo, Egypt,
in 2004 and 2008, respectively, both in electrical
engineering. He is currently pursuing the Ph.D.
degree in the Department of Electrical and Computer
Engineering, University of Waterloo, Waterloo, ON,
Canada.
His research interests include electrical machines,
reliability, renewable DG, distribution system planning, electric vehicles, storage systems, and smart
grid.

Yasser M. Atwa (S06) was born in Alexandria, Egypt, in 1975. He received the
B.Sc. and M.Sc. degrees in electrical engineering from Alexandria University,
Alexandria, Egypt, in 1998, and 2004, respectively, and the Ph.D. degree in
electrical engineering from the University of Waterloo, Waterloo, ON, Canada,
in 2010.
His research interests are distribution system planning, power quality, reliability, and distributed generation.

Ehab F. El-Saadany (SM05) was born in Cairo,

Egypt, in 1964. He received the B.Sc. and M.Sc. degrees in electrical engineering from Ain Shams University, Cairo, Egypt, in 1986 and 1990, respectively,
and the Ph.D. degree in electrical engineering from
the University of Waterloo, Waterloo, ON, Canada,
in 1998.
Currently, he is a Professor in the Department of
Electrical and Computer Engineering, University
of Waterloo. His research interests are distribution
system control and operation, power quality, distributed generation, power electronics, digital signal processing applications to
power systems, and mechatronics.